best-of-n-solving
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a workflow for spawning parallel subagents (
best-of-n-runner) to attempt different coding strategies concurrently. This creates a multi-step execution chain where subagent prompts are generated based on the primary agent's analysis of a problem. - Ingestion points: The process starts with a user or agent identifying a hard problem and defining strategies in
SKILL.md. - Boundary markers: The skill relies on the
best-of-n-runnersubagent type and git worktree isolation to separate execution environments. - Capability inventory: The runners perform file system operations (via git), code modifications, and test execution within their isolated branches.
- Sanitization: The methodology includes a comparison step (Step 3) where results are evaluated against success criteria before any changes are merged into the main branch, providing a validation checkpoint.
- [SAFE]: The use of isolated git worktrees for parallel experimentation is a security-positive pattern. It ensures that experimental code changes remain confined to specific branches and do not interfere with the primary codebase or other attempts until they are explicitly reviewed and merged.
Audit Metadata